How this instrument works
A point inside a loaded part rarely feels just one kind of stress. A gusset plate, a shaft fillet, a pressure-vessel wall — each carries some normal stress pulling along two directions plus shear coupling them, three numbers that do not individually say whether the material is about to yield. Von Mises stress collapses that trio into a single equivalent value, σᵥ = √(σx² − σxσy + σy² + 3τxy²), built so that comparing σᵥ against a uniaxial tensile yield strength gives the same yes-or-no answer a full multi-axis yield check would give. It is not a stress any gauge on the part actually reads; it is a bookkeeping number engineered to be comparable to one.
The shape of that formula comes from distortion-energy theory, published by Richard von Mises in 1913 after Tytus Huber described the same idea in a Polish journal in 1904 and Heinrich Hencky supplied the energy argument behind it in the 1920s. The theory splits the energy stored in a stressed element into a volume-changing part and a shape-changing part, and proposes that ductile metal yields once the shape-changing share alone reaches a critical value — not the volume-changing share, which a metal's tightly packed lattice tolerates almost indefinitely. Set σx and σy to zero and leave only τxy, and the formula reduces to σᵥ = τxy√3: pure shear needs only about 0.577 as much stress as pure tension to trigger the same yielding, which is exactly why fastener and shaft codes quote shear yield near 58 percent of tensile yield rather than half.
The criterion has a sharp boundary. It was built for ductile metals that yield by slipping along crystal planes, and it says nothing useful about cast iron, concrete, ceramics or bone, which fracture from the largest tensile stress present long before any distortion-energy threshold is reached — those materials want a maximum-normal-stress or Mohr-Coulomb check instead. Von Mises stress also throws away sign: a state in pure compression returns the identical σᵥ as the same magnitude in pure tension, so a buckling check or anything sensitive to hydrostatic pressure needs a different tool entirely.
- Enter the stress on the x-face into Normal stress, σx, in MPa — tension positive, compression negative.
- Enter the stress on the y-face into Normal stress, σy, keeping that same sign convention.
- Enter the stress coupling those two faces into Shear stress, τxy.
- Read Von Mises equivalent stress and divide your material's yield strength by it to get a factor of safety.
- For a plain uniaxial check, leave σy and τxy at zero — σᵥ then just equals σx.
Worked example — a pressure-vessel nozzle weld
A cylindrical pressure vessel carries 100 MPa of hoop stress in its wall, and thin-wall theory puts the axial stress at half that — but at this particular point, next to a nozzle weld, an FEA run instead reports 100 MPa hoop, 50 MPa axial, and 30 MPa of shear stirred in by the nozzle's own loading, a state close enough to what a real reinforcing-pad analysis returns. Enter 100 into Normal stress, σx, 50 into Normal stress, σy, and 30 into Shear stress, τxy. The formula runs σᵥ = √(100² − 100×50 + 50² + 3×30²) = √(10000 − 5000 + 2500 + 2700) = √10200 = 100.995 MPa.
Against a pressure-vessel steel with 250 MPa yield strength, that reading gives a factor of safety of 250 ⁄ 100.995 ≈ 2.48 — comfortably clear of yield, which is the number an inspector actually wants at that weld toe. Notice the shear alone contributed 2700 of the 10200 total under the square root, more than either normal-stress term on its own, which is the factor-of-3 weighting doing its work.
Questions
Why does shear stress get multiplied by 3 in the formula?
Because distortion-energy theory weighs shear more heavily than normal stress in the work needed to change a material's shape. Set σx and σy to zero and the formula reduces to σᵥ = τxy√3, so pure shear reaches the same equivalent stress as pure tension at only 1⁄√3 ≈ 0.577 of the tensile value — the textbook basis for quoting shear yield strength as roughly 58 percent of tensile yield rather than 50.
Can von Mises stress come out negative?
No. It is a square root of a sum of squared terms, so it is always zero or positive, regardless of whether the underlying stresses are tensile or compressive. That is a deliberate trade-off: the criterion discards sign entirely, so a state in pure compression returns the same σᵥ as the identical magnitude in pure tension, which is fine for predicting yield but useless for a buckling check.
Why did removing σy raise the answer instead of lowering it?
Because σy had been partly offsetting σx through the −σxσy term, and taking it out removes that offset. With σx = 100 MPa and τxy = 30 MPa held fixed, dropping σy from 50 MPa to 0 pushes σᵥ up from about 101.0 MPa to about 112.7 MPa. Adding a stress component does not always make the result worse, and removing one does not always make it better.
Is the von Mises criterion valid for cast iron or concrete?
No. It was derived for ductile metals that yield by slipping along crystal planes under distortion energy, and cast iron, concrete, ceramics and bone do not fail that way — they fracture from the largest tensile stress present, often at a flaw, well before any distortion-energy threshold matters. Those materials need a maximum-normal-stress or Mohr-Coulomb check, not this one.
How is factor of safety calculated from this result?
Divide the material's tensile yield strength by the von Mises stress: n = σyield ⁄ σᵥ. A steel part with 250 MPa yield subjected to a state that returns σᵥ = 100.995 MPa carries n ≈ 2.48, meaning the stress would need to grow roughly two and a half times over before that point starts to yield.
Does this replace a Mohr's circle or principal-stress calculation?
No, it answers a different question. Mohr's circle and principal stress locate the maximum normal and shear stress and the orientation they act at; von Mises stress collapses σx, σy and τxy into one scalar for a yield check and does not report a direction at all. Many analyses run both: principal stresses to understand the state, von Mises to judge it against yield.